PA

mihaiii_pallas-0.5-frankenmerge

PallasGGUF by TheBlokeUpdated 2024-01-06
Download on Hugging Face
Parameters36B
Context195K
Layers63
Architecturellama
Smallest quant15.3 GB
Quick answer

mihaiii_pallas-0.5-frankenmerge needs about 24.3 GB of memory at Q4_K_M (a 21.6 GB file) with an 8K context window, so it fits on a 32 GB GPU. This repository has 12 quants, from 15.3 GB (Q2_K) to 38.3 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 8.5 GB.

36B Pallas 0.5 Frankenmerge for reasoning and text comprehension

Mihaiii_pallas-0.5-frankenmerge is a 36B parameter model built on the Pallas family. It is a frankenmerge of Pallas‑0.5 and is optimized for long system prompts and reasoning tasks. The model is available in GGUF format and supports multiple quantizations.

Good at

  • Long prompt handling
  • Reasoning tasks
  • Text comprehension

Limits

  • Not suitable for story telling
  • No vision support
  • License is other

Best for

  • Chat assistants needing deep reasoning
  • Technical documentation summarization
  • Code review assistance

Who it’s for. Developers and researchers who need a large language model focused on reasoning and comprehension. The model is best used in local deployments with GPU acceleration.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K15.3 GB 17.9 GB
Too big Get file →
Q3_K_S15.7 GB 18.3 GB
Too big Get file →
Q3_K_M17.4 GB 20.1 GB
Too big Get file →
Q3_K_L19.0 GB 21.6 GB
Too big Get file →
Q4_020.4 GB 23.0 GB
Too big Get file →
Q4_K_S20.5 GB 23.1 GB
Too big Get file →
Q4_K_M21.6 GB 24.3 GB
Too big Get file →
Q5_024.9 GB 27.5 GB
Too big Get file →
Q5_K_S24.9 GB 27.5 GB
Too big Get file →
Q5_K_M25.5 GB 28.1 GB
Too big Get file →
Q6_K29.6 GB 32.2 GB
Too big Get file →
Q8_038.3 GB 40.9 GB
Too big Get file →

For 8 GB VRAM use Q2_K (file size 15.26 GB, memory 17.9 GB). For 12 GB VRAM use Q3_K_S (15.68 GB, 18.3 GB). For 24 GB VRAM use Q4_K_M (21.64 GB, 24.3 GB).

What GPU do you need for mihaiii_pallas-0.5-frankenmerge?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: no quant fits at an 8K context.
  • 6 GB VRAM: no quant fits at an 8K context.
  • 8 GB VRAM: no quant fits at an 8K context.
  • 12 GB VRAM: no quant fits at an 8K context.
  • 16 GB VRAM: no quant fits at an 8K context.
  • 24 GB VRAM: the best quant that fits is Q4_K_S (20.5 GB file, about 23.1 GB in use).
  • 32 GB VRAM: the best quant that fits is Q5_K_M (25.5 GB file, about 28.1 GB in use).
  • 48 GB VRAM: the best quant that fits is Q8_0 (38.3 GB file, about 40.9 GB in use).

Which mihaiii_pallas-0.5-frankenmerge quant should you choose?

Q4_K_M is the usual starting point for mihaiii_pallas-0.5-frankenmerge: it balances file size and quality well. Here it is a 21.6 GB file.

The Q8_0 file is 38.3 GB, about 1.8 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.

If memory is tight, the smallest file is Q2_K at 15.3 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturellama
Parameters36B
Layers63
Trained context200,000 tokens
KV cache (fp16)252 KB per token

Source

Quantized byTheBloke
FamilyPallas
Licenseother
Downloads (HF)1,971
Last updated2024-01-06

Frequently asked questions about mihaiii_pallas-0.5-frankenmerge

How much VRAM does mihaiii_pallas-0.5-frankenmerge need?

About 24.3 GB at Q4_K_M with an 8K context window. That is the 21.6 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run mihaiii_pallas-0.5-frankenmerge on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 17.9 GB, so you would need a larger GPU or CPU offloading.

What is the file size of mihaiii_pallas-0.5-frankenmerge GGUF?

It depends on the quant: Q2_K is 15.3 GB, Q3_K_S is 15.7 GB, Q3_K_M is 17.4 GB, Q3_K_L is 19.0 GB, Q4_0 is 20.4 GB, Q4_K_S is 20.5 GB and 6 more.

What context length does mihaiii_pallas-0.5-frankenmerge support?

The GGUF header lists a trained context length of 200,000 tokens (195K). The KV cache grows with the context you use: about 2.1 GB at 8K and 8.5 GB at 32K.

Where can I download mihaiii_pallas-0.5-frankenmerge GGUF?

The files are on Hugging Face in the TheBloke/Pallas-0.5-frankenmerge-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does mihaiii_pallas-0.5-frankenmerge use?

Hugging Face lists the license as other. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run this model on 8 GB VRAM?

Yes, using the Q2_K quantization which requires approximately 17.9 GB of VRAM at 8k context.

Does this model support vision tasks?

No, the model does not have vision capabilities.

What is the maximum context length?

The model was trained on 200000 tokens and supports up to that context length.

Other GGUFs of Pallas-0.5-frankenmerge

Similar models

Browse all →

More from TheBloke

All by TheBloke →

More Pallas models

All Pallas →

Other large models

All large models →